Συντάχθηκε 20-06-2012 10:47
από Galateia Malandraki
Email συντάκτη: gmalandraki<στο>tuc.gr
Ενημερώθηκε:
-
Ιδιότητα: υπάλληλος ΑΡΜΗΧ.
ΠΟΛΥΤΕΧΝΕΙΟ ΚΡΗΤΗΣ
Τμήμα Ηλεκτρονικών Μηχανικών & Μηχανικών Υπολογιστών
ΠΑΡΟΥΣΙΑΣΗ ΜΕΤΑΠΤΥΧΙΑΚΗΣ ΕΡΓΑΣΙΑΣ
Μπαλατσούκα-Στίμμινγ Αλέξιου-Κωνσταντίνου
με θέμα
“Σχεδιασμός LDPC Κωδίκων για το Κανάλι Πολλαπλής Πρόσβασης Δυο Χρηστών με Παρουσία Γκαουσιανού Θορύβου”
“Design of LDPC Codes for the Two-User Gaussian Multiple Access Channel”
Δευτέρα 25 Ιουνίου 2012, 10:30 πμ
Αμφιθέατρο, Κτίριο Επιστημών, Πολυτεχνειούπολη
Εξεταστική Επιτροπή
Καθ. Λιάβας Αθανάσιος (επιβλέπων)
Καθ. Σιδηρόπουλος Νικόλαος
Επ. Καθ. Καρυστινός Γεώργιος
Περίληψη
The capacity region of Gaussian Multiple Access Channels (GMACs) has been known since 1971. The efficient design of powerful codes that can achieve points near the dominant face of the capacity region, where the sum-rate is maximal, is an interesting problem. Significant progress has been made in this direction using time sharing, rate-splitting, as well as joint iterative decoding. Joint iterative decoding seems to be the most promising path, especially for codes that have low-complexity decoders, like Low-Density Parity-Check (LDPC) codes.
LDPC codes are capacity-approaching over a wide variety of channels. Additionally, elegant tools, such as Density Evolution and EXIT charts, can be used to accurately predict the asymptotic performance of an LDPC code ensemble. These tools can be used for the design of optimal LDPC codes, allowing for transmission over many types of channels with vanishingly small probability of error.
In this thesis, we focus on the two-user GMAC. To the best of our knowledge, there exist two LDPC code design frameworks for this channel. Amraoui et al. use Density Evolution, which is very demanding in terms of computational complexity. Roumy and Declercq use EXIT charts, a low-complexity approximation of Density Evolution, but their design is restricted to the case where the power of both users at the receiver is equal. We extend the EXIT chart based optimization framework by removing the equal power constraint, allowing for the optimization of LDPC codes over unequal power two-user GMACs. We show that, under some assumptions, the optimization problem can be expressed as an alternating linear programming problem, which can be solved efficiently. The resulting codes are close to optimal, in terms of sum-rate, and exhibit very good finite-length behavior.
Συνημμένα:
-
ΜετiErgasia-Anak-Balatsoukas.doc
Μεταφορτώσεις: 192,
Μέγεθος: 32 KB application/msword
